Reviewers now finish 1.7 times faster than in 2007, but it takes eight invitations to get two

The standing complaint about peer review is that reviewers are slow. A complete set of editorial records from one journal, published on 15 July 2026, says close to the opposite about the people who actually agree to do the work.

Paul A. Montagna, who has been a co-editor in chief of Estuaries and Coasts for more than nine years, downloaded the journal's editorial database covering 2 October 2007 to 31 December 2025 and analysed it. The database held 29,070 entries, each one a review invitation on a manuscript or on a revised version of a manuscript, and covered 4,393 manuscripts sent out for review across 18 years. The paper ran in Estuaries and Coasts, volume 49, article 155. Montagna stripped reviewer, author and editor names before analysis.

The recruitment numbers are the ones that have travelled. In 2007 an average of four invitations was needed to secure two reviews. By 2025 it took eight. The rate rose by an average of 3 per cent a year, and the linear trend accounts for 86 per cent of the variation. There was no meaningful difference in the number of invitations needed between papers that were eventually accepted and papers that were eventually rejected.

Underneath that sits a sharper change. Montagna separates people who decline from people who never reply at all, and it is the silence that has grown fastest. Between 2007 and 2010, on average fewer than one invited person per manuscript ignored the request. Since 2022 the figure has been more than two. Non-response has risen roughly fourfold since 2019, at an average of 8 per cent a year. Montagna describes non-responsiveness as the largest single change in reviewer behaviour over the period.

The finding that cuts against the popular account is what happened to the reviewers who said yes. In 2007 a first review took an average of 54 days to complete. In 2025 it took 32. That is a fall of about 41 per cent, and Montagna expresses it as reviewers completing their work 1.7 times faster in 2025 than in 2007. The trend was noisy before 2015 and steadier afterwards. The decline averaged about 2 per cent a year.

That distinction matters because it points different reforms in different directions. If reviewers were the slow part, then shorter deadlines, payment for reviews or automated chasing would be the obvious levers. If the slow part is getting anyone to accept in the first place, those levers do comparatively little, and the problem becomes one of recruitment, workload and recognition.

The rest of the timing data explains where the months actually go. Montagna reports that 71 per cent of the initial decision period is consumed by finding reviewers and completing reviews. An initial review cycle runs about 80 days at the median, with a quartile range of 55 to 120 days. Manuscripts that need revision before acceptance take roughly 160 to 200 days, because the cycle simply repeats. Final decisions split sharply by outcome: rejection came at a median of 102 days and a mean of 169, while acceptance came at a median of 223 days and a mean of 250. Rejections are quicker largely because most rejected manuscripts were reviewed only once.

The paper is not internally consistent on one figure. Its results section gives the decline in final decision time as 1 per cent a year for accepted manuscripts and 4 per cent a year for rejected ones. Its discussion section reverses the emphasis, giving 1 per cent a year for rejected papers and 2 per cent for accepted ones. The direction of travel is the same in both passages, but the rates are not, and the paper does not reconcile them.

Montagna's data comes from a single journal in a single field, which is the obvious limit on it. Two independent datasets point the same way on recruitment. Silverchair, the scholarly publishing technology company, reported in 2026 on eight years of activity on its ScholarOne Manuscripts platform covering 2018 to 2025, alongside responses from more than 2,000 reviewers, authors and editors. It found that editors now need an average of 4.5 invitations to secure a single review, close to double the 2018 rate, and that for every 100 invitations sent, editors wait a combined 407 days on people who eventually decline or never answer. That is a company reporting on data from its own platform.

Separately, Hamid Jamali, Edward Luca and Simon Wakeling of Charles Sturt University surveyed 139 journal editors and interviewed 27 of them, writing the work up on 16 February 2026. They found 55 per cent rated finding reviewers a significant challenge, with some editors reporting more than 30 invitations to secure two reviewers. They argue the cause is that reviewing is unpaid, unrecognised and increasingly absent from university workload models.

What none of these datasets settles is why individuals stop answering. Montagna's records show that invitations go unanswered, not why. Whether the cause is volume of requests, irrelevance of the requests, email filtering, or a withdrawal of goodwill is not visible in editorial metadata, and the interventions that follow depend on which it is.